Stolen Fireworks AI Key Led to Trojanized Claude Code Extension on Cursor Marketplace

Cursor disclosed that a stolen access key exposed by Fireworks AI enabled attackers to trojanize Anthropic's Claude Code extension on its marketplace, potentially affecting up to 5,000 users who downloaded the malicious build on October 29, 2026.

Christopher Ort
By Christopher Ort · 3 min read
Share
A developer's laptop screen showing a code editor with a warning symbol, photographed from over the shoulder in a dimly lit office
Image: AI-generated, i10X
On this page
  1. Quick Take
  2. In plain English
  3. What happened
  4. Broader impact of Fireworks AI breach
  5. Sources
i10X Quick Take

Quick Take

  1. What happened: Attackers stole a Fireworks AI access key from Cursor’s read-only storage bucket, used it to access internal systems, and trojanized Anthropic’s Claude Code extension on Cursor Marketplace on October 29, 2026.
  2. Who is affected: Up to 5,000 Cursor users who downloaded Claude Code extension versions 2.1.312 or 2.1.313 during an 11-hour window on October 29, 2026.
  3. What to do: Affected users who installed the malicious versions have been notified by Cursor and should revoke exposed credentials.
  4. Current status: The malicious extension was removed within 11 hours, the Fireworks AI key was deactivated, and no additional malicious activity has been detected since November 1, 2026.
  5. Broader impact: The Fireworks AI breach also affected Uber (thousands of driver’s licenses stolen) and Notion.

In plain English

Cursor, a company that makes AI-powered coding tools, discovered that hackers stole a security key that Cursor had given to a partner company called Fireworks AI. The hackers used that key to break into Cursor’s systems and replace a legitimate coding extension (a plugin called Claude Code made by Anthropic) with a fake, malicious version. About 5,000 people might have downloaded the bad version on October 29, 2026, before Cursor caught it and took it down 11 hours later.

What you should do: If you use Cursor and downloaded that extension on that day, the company has already contacted you to help fix the problem. If you were not contacted, you do not need to do anything. The same Fireworks AI breach also hit Uber and Notion, so this was a bigger supply-chain attack affecting multiple companies.

What happened

Cursor disclosed on Friday that the compromise of Anthropic’s official Claude Code extension on its marketplace stemmed from the theft of an access key the company had issued to Fireworks AI. According to Cursor, the key was exposed by Fireworks AI, and attackers used it to copy data from a read-only storage bucket Cursor used to share data with the vendor.

The read-only bucket contained limited coding agent traces from sessions that were not in Privacy Mode, a Cursor setting that lets users opt out of data being used for training. Most of these sessions dated back to 2025. Among the traces were credentials that had not been edited out, exposing the API credentials of less than 0.01 percent of customers.

Attackers used these credentials to reach select software-as-a-service tools used by Cursor and accessed the following data:

  • Select archived non-production Cursor source code repositories, including some containing business contact information
  • Limited customer support data and email conversations
  • Product usage metrics and contact details
  • Summaries of some coding agent conversations
  • Some credentials users had sent to the Cursor support team

Cursor said it deactivated the vendor’s key, revoked and rotated credentials, and hunted for signs of deeper access after identifying further access. The company temporarily disabled internal and product services, disrupting access to some Cursor features.

On October 29, 2026, the attacker used a stolen credential from the exposed data to place trojanized versions of Anthropic’s Claude Code extension directly on the cache used to serve Cursor’s extension marketplace. At most 5,000 users who requested specific cached versions of the extension on October 29 potentially received the malicious build, which was available for approximately 11 hours before Cursor removed it.

Cursor took the extension marketplace offline and worked with the cloud service provider hosting the attacker’s server to take it down. The company notified affected users who installed versions 2.1.312 or 2.1.313 from the marketplace during the exposure window and helped revoke exposed credentials.

Cursor has moved all inference traffic off Fireworks AI, deactivated the key issued to it, and cut off its access to all data sets shared with it. The company emphasized that no additional malicious activity has been detected since November 1, 2026.

The incident did not impact users who had enabled Privacy Mode, customer code repositories, Cursor’s production environment, the Cursor app and its official updates, other apps, or any other extension or plugin downloaded from Cursor Marketplace, according to the company.

Broader impact of Fireworks AI breach

Uber confirmed that unknown threat actors stole thousands of drivers’ license information after obtaining internal credentials. On October 9, 2026, a service provider notified Uber that an unauthorized third party had compromised the service provider’s internal credentials and used them to access Uber’s account with the service provider. On October 13, 2026, Uber’s investigation determined that the unauthorized party had used the compromised credentials to access Uber data connected to the service provider.

Although Uber did not disclose the name of the provider, TechCrunch reported that the provider in question is Fireworks AI. Uber has not disclosed how many users were impacted by the breach.

Last week, online productivity startup Notion also confirmed its data was stolen in the Fireworks AI hack.

Sources

Christopher Ort

Christopher Ort

Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). He is the author of “The Style Penalty”, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.

Editor's picks

  1. CVE-2026-35029: LiteLLM Gateway Authorization Bypass Enables RCE
  2. Cursor marketplace breach traced to unredacted credentials in stored agent traces
  3. AI-Powered Antidetect Browsers Enable Mass Social Media Fraud and Influence Operations
  4. Aokigahara Infostealer Targets AI Tools and Developer Credentials
  5. Yoba malware campaign abuses signed ChatGPT and Claude binaries via DLL sideloading